Blizzard began experimenting with machine learning to automatically detect toxic language and behavior in Overwatch, without waiting for players to submit reports. Game Director Jeff Kaplan publicly confirmed these experiments in April 2018. The current operational status of a dedicated ML moderation system for Overwatch specifically — separate from the ToxMod system used in Call of Duty — is not fully documented.

Kaplan described experiments in which Blizzard was training AI to recognize toxic language in context, noting that the system could identify problems without requiring a human complaint. He also described aspirations to detect behavioral toxicity — such as players deliberately sabotaging teammates — rather than just offensive speech. Blizzard joined a cross-industry anti-toxicity initiative around the same time. A 2024 academic study using publicly available Overwatch voice chat data built an external toxicity prediction model that achieved 86.3% accuracy using machine learning, suggesting the data signals are detectable. Whether Blizzard currently operates its own ML moderation layer inside Overwatch 2, or relies on ToxMod (documented in Practice 2), is not publicly confirmed.